I keep circling a belief I can’t fully explain yet: the thing we build should be usable by agents as much as by humans. Most software assumes a person is the one clicking, reading, deciding. But if agents are going to be doing real work inside these tools, that’s a different design target. The human-facing surface and the agent-facing surface might not be the same thing at all.
I’m honest that I don’t yet understand all of what this means in practice. There’s a word people keep using, primitives, and I want to actually understand it instead of nodding along. My instinct is that getting the underlying building blocks right matters more than the screens on top, because that’s what an agent reaches for.
If you were designing a product today, would you build it for the human or for the agent doing the work, and are those even the same design?